I made a mess the first time with this, so forgive me for doing it again...please.
Here is Freeman's revised text:
It's a pdf file.
Part 2 is coming, he says.
In Part 1 (above link) he has revised his figure down to 662,000 to 1 against the exit poll discrepancy for the three states considered being due to random chance or error. Previously he had it at 250 million to 1.
He then says that as much as one can be sure of anything in social science, we can be sure that the discrepancy was not due to chance or random error.
In Part 2 (not included in the above link), he is going to attempt an answer to the question, what caused the discrepancy.
Because I think it's pretty important, and Part 2 will deal with the question of why the exit polls were off with new analysis added, I'm highlighting Freeman's ongoing work by giving it a new diary.